This proposal seeks support to continue the work on the application of statistical discrimination methods using log linear contingency models in the development of automated record linkage systems. The record linkage problem is one of matching with minimal possible error the records of an individual in one computer file with his records in another. Efficient linkage rules are of central importance in many areas of health services research. The application of these statistical techniques is of value in designing and carrying out epidemiological and administrative studies which require the matching of large data files. The techniques developed are useful in many problems that require the analysis of qualitative data. The proposal seeks continued support to 1) investigate how recent advances in the applications of log linear models can be applied to the record linkage problem, 2) complete the development of computer programs needed to design linkage rules, 3) analyze the data from the record linkage system of Strong Memorial Hospital in order to design rules to test the statistical methodology, and 4) if feasible, do preliminary design on a computer system to be used administratively for record linkage of patients admitted to Strong Memorial Hospital. The first phase of this study has been directed at the development of statistical methodology. The work has been substantially completed. In this second phase of the study, we seek support 1) for development of data management programs to evaluate a number of different linkage rules for the Strong Memorial Hospital data, 2) for development of accurate estimates of the error to be expected in hospital applications, 3) to study applicatons on a realistic example based on Strong Memorial Hospital data, 4) to investigate applications of models for incomplete contigency tables which may improve the precision of discriminators in hospital applications, 5) to investigate the effects of assumptions about cost of misclassification of the discriminators, 6) to develop computer programs and statistical analyses so that they can be easily carried out at other hospitals, and 7) to continue to investigate the feasibility of developing such a system for Strong Memorial Hospital.